You are in Ibiza, where babes of all around the world gather to display their tight bodies and trimmed honeypots. Pack your telephoto, plenty of water and enter the ultimate voyeurism heaven!
This Lora realistically captures the style of beach voyeur telephoto, which is a combination of:
Camera angle - either a ground-level or eye-level shot with telephoto.
Dynamic composition - rich scenes featuring intrincate depth of field, with elements in the foreground and background.
Slice of Life - unaware women acting and posing in a natural way.
-> Get the best results with deis_3m + beta57.
Training keywords
Basic camera setup: front view, side view and rear view. In combination with optional closeup.
Foreground: ground is blocking the photo in the foreground will give you the foreground depth of field. You can experiment with someone, bush, tree leaves or whatever object in the foreground.
Poses: standing, sitting and laying.
Single or multiple subjects: a woman or two women next to each other.
You can combine the keywords above as you see fit in addition to the usual keywords for describing how the women should look like, if they are looking at viewer, etc ...
What to avoid
By just changing the seed it will generate great varied results on this particular theme, but there is one thing you need to be aware of:
It's a style! Do not override or misguide it with your prompt (unless you are exprimenting of course!). That means that, if you want to generate over the original behaviour, avoid specifying color palettes, extreme camera angles, bokeh, etc ...
Enjoy
Leave a like and comment because, well, that let's me know you've enjoyed it.
Description
FAQ
Comments (11)
I tried using it, but all it seems to do is put a visible checkerboard pattern over the image and make the quality overall a bit odd.
Edit: it does not interact well with character loras, this is the cause.
Odd. How did you try?
me too, what a shame...it got so much potential!
Regarding your mention to not working with character Loras, I see, didn’t really try that.
@oo4766229 is it because it’s conflicting with character Loras as well?
@sparkfun Yeah. Even tried to lower the weight to 0.35, still no luck. And it breaks the likeness greatly. Don't think it would be easy to train a concept on Chroma though.
@sparkfun yes it also make character identity shift as well.
@TheNecr0mancer I’m training an experimental version tweaking some parameters to see if it reduces the overlap with other Loras.
@sparkfun looking forward to it.
i was able to sort of work around the line issue by combining this lora with the lenovo lora and using res_2m/beta57, but even then lines and bad quality was not completely gone, with a person detailer and upscaler you can kind of clean up the bad behavior after the fact though.
Works very well👍🏻









